Anonymization of distribution feeder data using statistical distribution and parameter estimation approach. (August 2022)

• Record Type: Journal Article
• Authors: Ali, Muhammad; Prakash, Krishneel; Macana, Carlos; Rabiul, Md; Hussain, Akhtar; Pota, Hemanshu
• Is Part Of: Sustainable energy technologies and assessments. Volume 52:Part C(2022)
• Publication Date: 2022-08
• Subjects: Data anonymization -- Data analytics and protection -- Power distribution networks -- Power distribution feeders -- Statistical distributions; Renewable energy sources -- Periodicals; Energy development -- Technological innovations -- Periodicals; Electric power production -- Periodicals; Energy storage -- Periodicals; 333.79
• DOI: 10.1016/j.seta.2022.102152
• Languages: English
• ISSNs: 2213-1388

Abstract

Abstract: Distribution networks are undergoing a transition due to the rapid increase in distributed energy resources (DERs). Concerns over revealing critical infrastructure information and breaching consumer privacy have significantly hindered their development. Many applications that require data sharing are still impractical due to the lack of sufficient datasets, privacy concerns, and data access problems. To address these issues, a statistical distribution approach for parameter estimation is proposed. The statistical patterns of real distribution feeders are examined by accessing the confidential database of a local distribution network service provider (DNSP). An algorithm based on the Maximum Likelihood Estimate (MLE) is applied to estimate the statistical distribution parameters that represent the actual data. Then, these statistical distribution parameters are used to generate anonymized datasets that are realistic. A Kolmogorov–Smirnov (K-S) test is conducted to confirm the effectiveness of anonymized datasets, and results are compared with the actual feeder datasets. Validation is carried out with existing methods and comparisons are shown on the different portions of the datasets (25 percent, 50 percent, 75 percent, and 100 percent). The comparison results indicate superior performance over traditional methods, with a performance improvement ranging from 1 to 13 percent. The practical application of the method is demonstrated on the IEEE 123-node test feeder. The method achieves consistent results on voltage profiles, with a maximum difference of 0.420 percent between actual and anonymized datasets.